SuperMemo's creator built a learning algorithm that won't let you forget

The Problem on a Page of Word Pairs

In the summer of 1985, a twenty-three-year-old graduate student in Poznań, Poland sat down with pages of roughly forty English word pairs and ran a quiet experiment on himself. He was not trying to build a software company. He was trying to stop forgetting. Piotr Woźniak had a problem familiar to every language learner: the vocabulary he collected so carefully kept evaporating. He had been studying English since secondary school, and by then he was also deep in a degree in molecular biology at Adam Mickiewicz University. The words he memorized for class would fade within days. He would return to them and find only fog where certainty used to be.

So he measured. Working with those pages of word pairs, he tested how long the gap before a review could grow before recall broke down. He tracked his own forgetting curve with the patience of a bench scientist. The finding that emerged from that summer was simple but consequential: after each successful review, the next interval could be made considerably longer without increasing the risk of forgetting. Review after one day. Then seven. Then sixteen. Then thirty-five. The intervals roughly doubled. He called the paper-and-pencil procedure Algorithm SM-0, and he studied with it for two years before he ever touched a computer.

That patient self-experiment, conducted alone in a Polish university town in the mid-1980s, would eventually become the foundation of a learning system used by millions of people around the world. The software that grew from it SuperMemo, created by Woźniak and released for MS-DOS in December 1987 is widely regarded as the first spaced repetition program that actually worked. Not a flashcard display. A memory model. A system that calculated, for every single item a learner studied, the optimal moment for the next review, based on a model of human forgetting.

From Paper to PC: The Amstrad That Changed Everything

In 1987, Woźniak got access to his first personal computer an Amstrad PC 1512. In December of that year, he wrote the first computer version of his system: SuperMemo 1.0 for DOS. The move from paper to software changed something fundamental about what was possible. With paper, you could only schedule whole pages of vocabulary at once. With software, the program could schedule every item individually, according to how well that particular item was actually remembered.

This was the shift that mattered. The algorithm did not just remind you to study. It watched how you performed and adjusted accordingly. Each flashcard became its own little experiment in forgetting and recovery, with its own schedule derived from your own performance history. The system learned you as you learned through it.

The algorithm carried by those early DOS versions described in detail in Woźniak's 1990 master's thesis Optimization of Learning and later published openly on the SuperMemo website became known as Algorithm SM-2. It is disarmingly simple in structure, even as its implications are wide-ranging. Every item gets its own "easiness factor," a number that starts at 2.5. After each review, the learner grades their own recall on a scale from 0 to 5, and that grade nudges the easiness factor up or down. The first interval is one day. The second is six days. After that, each new interval is the previous one multiplied by the item's easiness factor. An item that is failed goes back to the beginning of the cycle and returns quickly.

In human terms: intervals grow roughly exponentially. Easy material drifts far apart. Difficult material keeps coming back. Every flashcard ends up with a personal review schedule. That is the whole trick and it works, because it approximates the natural rhythm of memory consolidation and decay.

What the Algorithm Actually Does

The SM-2 algorithm operates on a fundamental principle of cognitive science that psychologists had been describing since the 1930s: the spacing effect. Your brain retains information better when you review it at increasing intervals over time, more than cramming everything at once. What Woźniak did was turn that principle into a computational procedure that could be automated.

The process works like this. When you review a flashcard, you rate how easy or hard it was to recall the information on a six-point scale. A 5 means you recalled it with perfect ease. A 4 means correct after hesitation. A 3 means correct but with serious difficulty. A 2 means you got it wrong but felt familiar with the answer. A 1 means you got it wrong but recognized the answer when you saw it. A 0 means complete blackout no recollection whatsoever.

That rating feeds back into the algorithm. If you rate a card as easy (4 or above), the easiness factor for that item increases slightly, and the next interval grows longer. If you struggle (below 3), the easiness factor decreases, and the card comes back sooner. The minimum easiness factor is 1.3, which prevents the system from giving up on difficult material entirely.

The beauty of the system is that it adapts to your individual performance without any external intervention. Difficult cards appear more frequently. Easy cards drift far apart. You do not have to decide when to review the algorithm decides for each item based on your own demonstrated recall. Over time, the system builds a personalized map of your memory for every piece of knowledge you enter.

This is why the algorithm became so influential. It was not just a flashcard program. It was a memory model the first practical answer to the forgetting curve described by Hermann Ebbinghaus in 1885 that could actually run on a personal computer and scale to thousands of items.

The Open Algorithm and the Industry It Built

In 1991, Woźniak and Krzysztof Biedalak co-founded SuperMemo World in Poland, becoming the first company in the world to use spaced repetition commercially to aid learners. The company began offering courses and software, and the SuperMemo method spread. But what made the algorithm truly transformative was not the company it was the openness with which Woźniak published the details.

The SM-2 algorithm was published openly and became the de facto standard of an entire industry. It still runs, in adapted forms, inside Anki, Mnemosyne, and countless other flashcard applications. A 2025 analysis by the learning platform Tegaru notes that SM-2 "laid the foundation for every modern SRS tool" and that its "simplicity, effectiveness, and open-source nature have made it the gold standard for spaced repetition systems."

"SM-2 introduced a scientific, data-driven approach that schedules reviews right before you're likely to forget, resulting in 200-300% better retention compared to traditional study methods."

That quote comes from Tegaru's explanation of the SM-2 algorithm, which documents how the system works and why it outperforms conventional study methods. The platform notes that SM-2 powers hundreds of learning applications today, including Anki, Quizlet, Memrise, and Tegaru itself though many modern implementations have since developed their own algorithms that build on Woźniak's original framework.

The ripple effects of that openness are hard to overstate. A single researcher in Poland, publishing his algorithm in the late 1980s and early 1990s, set the technical standard that an entire global industry would adopt and adapt. He did not patent the core idea. He wrote about it in academic papers, described it in his master's thesis, and published the details on the SuperMemo website. The learning world absorbed them and built from there.

Four Decades of Refinement

Woźniak did not stop after SM-2. His work continued through the 1990s and 2000s as he refined the algorithm, expanded the software, and deepened his research into the science of learning and memory.

In 1995, he received a doctorate from the Wrocław University of Economics. His doctoral dissertation was entitled Economics of Learning: New Aspects in Designing Modern Computer Aided Self-Instruction Systems. The title captures something important about his approach: he was not just building software. He was thinking about the economic efficiency of learning the relationship between effort spent and knowledge retained, and how software could optimize that ratio.

He served as President and Head of R&D at SuperMemo World from 1991 to 1997, and in 1997 founded SuperMemo Research, an independent R&D unit that continues his work today. The desktop software evolved through multiple versions, adding features like incremental reading in SuperMemo 10 (2000), which allowed users to import longer texts and process them in segments scheduled for review over time. He later developed the concept of neural creativity, first implemented in SuperMemo 17 (2016).

His research interests expanded well beyond the original algorithm. According to his profile on supermemo.guru, he has written extensively about the failure of schooling, the optimization of sleep, the learn drive, and the molecular and neural correlates of long-term memory. He has published on topics ranging from dyslexia and coercion to natural creativity cycles and the impact of education on learning and creativity.

He prefers anonymity, he has said, because it allows him to focus on his learning without distraction. In a field where visibility often drives funding and influence, he has largely stayed out of the spotlight, letting the algorithm speak for itself.

The Learn Drive and the Critique of Schooling

One of the threads that runs through Woźniak's later work is a sustained critique of how formal education is structured. He believes that learning needs to be driven by the natural "learn drive" the intrinsic motivation to explore, understand, and remember that humans are born with. He has written extensively about what he calls the "problem of schooling," arguing that coercive educational systems work against the brain's natural learning mechanisms.

This critique is rooted in his research on memory, sleep, and adaptability. If the brain has evolved powerful systems for deciding what to learn, when to review it, and how to consolidate it during sleep, then forcing learners into arbitrary schedules and arbitrary content makes those systems work less effectively, not more. The learn drive, in his framing, is not a soft metaphor. It is a measurable biological tendency that education policy largely ignores.

His work on sleep optimization reflects the same integrative approach. He has studied how memory consolidation happens during sleep, how homeostatic and circadian processes interact, and how learning can be optimized in relation to the sleep-wake cycle. These are not peripheral interests. They are central to his understanding of what learning actually is a biological process that software can support but never replace.

Why This Matters for EducationGuide Readers

For readers researching learning systems, frameworks, and the people behind them, Woźniak's story offers a particular kind of value. It is not the story of a company that scaled, though SuperMemo World did grow. It is the story of a single person who identified a problem in his own learning, measured it with scientific rigor, built a solution that worked, and then published it openly so that others could build further.

The practical payoff is concrete. If you use Anki, Quizlet, Memrise, or any other spaced repetition flashcard system, you are using software that traces its lineage to Algorithm SM-2 and the experiments Woźniak ran on himself in Poznań in 1985. The reason those systems work the way they do the reason they schedule reviews individually, adjust intervals based on your performance, and prioritize difficult material is because of the framework he developed. Understanding that origin does not just satisfy curiosity. It gives you a clearer picture of what the tools are actually doing and why.

For educators, trainers, and anyone designing learning experiences, the SuperMemo story is also a case study in what it looks like when a learning system is built around the actual mechanics of memory more than around institutional convenience. Woźniak's research into the learn drive and his critique of schooling are not just philosophical positions. They are empirically grounded arguments about what makes learning efficient, and they have direct implications for how courses, curricula, and self-study programs can be designed better.

The Algorithm Today

SuperMemo continues to be developed. According to Wikipedia's entry on SuperMemo, the stable release as of June 29, 2026 was Version 20.00.32. The software has come a long way from the DOS version Woźniak wrote on an Amstrad PC 1512 in 1987. It now supports images, video, HTML questions and answers, and incremental reading workflows that let users process entire articles or books inside the program.

But the core insight remains the same. The algorithm still schedules each item individually based on your demonstrated recall. The easiness factor still adjusts after every review. The intervals still grow exponentially for well-remembered material and contract for difficult material. The fundamental architecture of the system the thing that makes it work is still the architecture Woźniak designed in his 1990 master's thesis.

Modern alternatives like FSRS (Free Spaced Repetition Scheduler) have been developed to address some of SM-2's limitations, and platforms like Anki and Tegaru have moved to these newer algorithms. But SM-2 remains the foundational reference point. It is the algorithm that proved the concept, established the standard, and showed that a simple mathematical model of forgetting could be turned into a practical tool that millions of people use every day.

What the Numbers Cannot Capture

There is something worth pausing on in the story of how SuperMemo began. Woźniak did not start with a business plan. He did not start with a grant or an institutional mandate. He started with a personal frustration the feeling that his carefully collected English vocabulary was slipping away and he responded by measuring that frustration with the tools he had available. Paper. Pencil. A stack of word pairs. The patience to run the same test again and again until he understood what was happening in his own memory.

That approach starting with your own experience of a problem, measuring it carefully, and building a solution from the data is a model for how learning systems can be designed. It is also, not coincidentally, the approach that cognitive science itself recommends: base your methods on evidence about how the mind actually works, not on assumptions about what ought to work.

Woźniak has spent four decades refining that approach. The algorithm has grown more sophisticated. The software has added features. The research has expanded into sleep, creativity, dyslexia, and the molecular biology of memory. But the starting point has not changed. It is still the learner, sitting with material they want to remember, trying to find the rhythm that makes retention sustainable.

Where to Read Further

For readers who want to go deeper into the history, science, and current state of spaced repetition and the SuperMemo method, the primary sources offer a rich trail.

• SuperMemo: Piotr Woźniak and the first app that really works (1990) A detailed account of the 1985 experiments, the move from paper to PC, and the development of Algorithm SM-2, written for a general audience interested in the history of educational technology.
• The SuperMemo method The official description of how the method works, with practical guidance on using the software and understanding the principles behind spaced repetition scheduling.
• Piotr Wozniak on supermemo.guru Woźniak's own platform, offering his current research interests, publications, and perspective on learning, memory, education, and the learn drive.
• SM-2 Algorithm Explained: The Science Behind Spaced Repetition A clear technical walkthrough of how the algorithm calculates intervals, adjusts easiness factors, and schedules reviews, with context on its enduring influence on modern learning tools.

A Quiet Legacy

The story of Piotr Woźniak and SuperMemo is not a story about disruption or scale or venture capital. It is a story about a person who paid close attention to his own forgetting, ran careful experiments to understand it, and built a system that has helped millions of people remember what they want to remember. The algorithm he published openly in 1990 became the technical foundation for an industry. The philosophy of learning he developed over four decades rooted in the learn drive, in the science of sleep, in the critique of coercive schooling offers a coherent alternative to how most educational systems are currently designed.

He prefers anonymity. He works on his own terms. He lets the algorithm do the talking. And the algorithm, after all this time, still speaks clearly: review what you are about to forget, and you will remember it longer. That is the whole trick. And it works.